Design method of corrosion-resistant multi-principal-element alloy coating based on machine learning model

Abstract

The invention discloses a method for designing a corrosion-resistant multi-principal-element alloy coating based on a machine learning model, and relates to the technical field of multi-principal-element alloy coatings, and the method comprises the following steps: S1, constructing a data set and preprocessing the data set, the data set comprising alloy components, magnetron sputtering process parameters and electrochemical corrosion performance indexes of the coating, dividing the preprocessed data set into a training set and a test set; s2, performing fitting, training and modeling by using various classic machine learning models to obtain corresponding target prediction models, and evaluating the target prediction models based on the test set to select a final target prediction model; and S3, on the basis of the final target prediction model, a genetic algorithm is used for searching an optimal solution in a multi-target space so as to determine the optimal element content and magnetron sputtering process parameters, and therefore the multi-principal-element alloy coating is prepared. According to the method, the component-process-performance integrated design can be realized, the alloy design and the preparation process of the corrosion-resistant multi-principal element alloy coating are guided, and the development period of a new material is shortened.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-principal component alloy coatings, and in particular to a design method for corrosion-resistant multi-principal component alloy coatings based on a machine learning model. Background Art

[0002] As advanced metal materials are increasingly used in harsh industrial environments, especially in the oil, marine, nuclear and energy sectors, corrosion has become a serious challenge. Globally, corrosion causes economic losses of more than $4 trillion each year. Corrosion not only shortens the service life of metal parts, but also poses huge safety risks. Pitting corrosion is particularly worrying because it damages the integrity and service life of the material structure and may lead to complete failure of metal parts. Therefore, the development of advanced protection solutions has become increasingly important. At present, the preparation of multi-principal alloy coatings has become a promising method that can provide enhanced corrosion resistance and durability in harsh environments. Therefore, exploring multi-principal alloy coatings is of great significance in addressing the challenges posed by corrosion.

[0003] The design strategy of multi-principal alloys provides a broader composition space for the development of new materials with excellent properties, overcoming the single principal element design limitation of traditional alloys. Considering the high entropy effect, multi-principal alloys are more likely to form single-phase solid solutions, creating inherent favorable conditions for designing corrosion-resistant materials. In addition, unlike traditional stainless steels, the composition space of multi-principal alloys can contain high percentages of passivating elements such as Cr, Ti, Al, and Ni, thereby providing higher corrosion resistance. However, this composition design has the disadvantages of high cost and industrial application limitations. Compared with bulk multi-principal alloys, the microstructure and properties of multi-principal alloy coatings are easier to adjust, thus saving material costs. Most notably, magnetron sputtering technology has promoted the high-throughput development of multi-principal alloy coatings, significantly accelerating the design and application of new materials. Multi-principal alloy coatings retain all the favorable properties of bulk multi-principal alloys, such as slow diffusion and simple phase structure, which contributes to the corrosion resistance of the coating. Therefore, the development of corrosion-resistant coatings is crucial to improving the service life of metal materials in the marine, shipbuilding, and petroleum fields.

[0004] The traditional trial and error method is currently the main method for developing new multi-principal alloy coatings, but this process may require a lot of labor and resources. With the introduction of the Materials Genome Project, machine learning and data-driven methods have been applied to the rapid design of various advanced materials, greatly shortening the design cycle and cost of new materials. Through data mining, machine learning can autonomously identify the complex nonlinear relationship between the required physical properties and the target variables. At present, machine learning has been applied to the field of coatings, but the research on the performance of multi-principal alloy coatings is limited, and mainly the auxiliary design of coating hardness and wear resistance has been carried out.

[0005] Xu et al. (X.Xu, X.Wang, S.Wu, L.Yan, T.Guo, K.Gao, X.Pang, AAVolinsky, Design of super-hard high-entropy ceramics coatings via machine learning, Ceram. Int. 48 (2022) 32064-32072.) used machine learning algorithms combined with high-throughput experimental methods to rapidly develop a new high-entropy ceramic coating (AlCrNbTaTi)N with a hardness of 40.1 GPa, an increase of 9% over the original five-element system. In addition, Wu et al. (S. Wu, X. Xu, S. Yang, J. Qiu, AAVolinsky, X. Pang, Data-driven optimization of hardness and toughness of high-entropy nitride coatings, Ceram. Int. 49 (2023) 21561-21569.) developed a high entropy nitride coating with an optimized combination of hardness and elastic modulus by using a machine learning method with multi-objective optimization. Jia et al. (B. Jia, Q. Wan, L. Yan, Y. Luo, Q. Wei, C. Niu, B. Yang, S. Li, L. Meng, Tribological properties and machine learning prediction of FeCoCrNiAlN high entropy coatings, Surf. Coat. Technol. 477 (2024) 130341.) used an extreme value gradient enhancement regression algorithm to predict the friction coefficient of FeCoCrNiAlN coatings.

[0006] However, machine learning has not yet been used to achieve the integrated design of corrosion resistance and preparation process of multi-principal alloy coatings, which is crucial for the rapid development of multi-principal alloy coatings with high corrosion resistance. Therefore, it is urgent to develop a design scheme for corrosion-resistant multi-principal alloy coatings based on machine learning. Summary of the invention

[0007] The purpose of the present invention is to provide a design method for corrosion-resistant multi-principal alloy coatings based on a machine learning model, combine the machine learning algorithm with magnetron sputtering technology, guide the design and preparation process of corrosion-resistant multi-principal alloy coatings, and shorten the development cycle of new materials.

[0008] To achieve the above object, the present invention provides the following solutions:

[0009] A design method for a corrosion-resistant multi-principal alloy coating based on a machine learning model, comprising:

[0010] S1, constructing a corrosion data set of a multi-principal alloy coating in 3.5wt.% NaCl and preprocessing it, wherein the corrosion data set includes alloy composition, magnetron sputtering process parameters and electrochemical corrosion performance indicators of the coating, and the preprocessed corrosion data set is divided into a training set and a test set;

[0011] S2, in Python's scikit-learn library, uses a variety of classic machine learning models to fit and model the training set, and obtains multiple target prediction models after training; then uses the test set to evaluate these target prediction models, and selects the model with the best performance as the final target prediction model;

[0012] S3, based on the final target prediction model, use a genetic algorithm to search for the optimal solution in the multi-target space to determine the optimal element content and magnetron sputtering process parameters, and prepare a multi-principal alloy coating according to the optimal element content and process parameters.

[0013] Furthermore, the preprocessing includes: performing a Pearson correlation coefficient analysis on the corrosion data set and drawing a heat map to evaluate the correlation between component characteristics and process parameter characteristics, and combining feature importance sorting to eliminate features with a correlation coefficient greater than 0.9.

[0014] Furthermore, in S2, multiple target prediction models are obtained through training, specifically including: based on the training set, taking the composition ratio of alloy elements and process parameters as input variables, and taking the electrochemical corrosion performance index as output variable, the selected target prediction model is trained to obtain a trained target prediction model.

[0015] Furthermore, there are five classic machine learning models, including random forest, gradient boosting regression, artificial neural network, support vector regression and extreme gradient boosting model.

[0016] Furthermore, the magnetron sputtering process parameters include sputtering power, bias voltage, substrate temperature and sputtering time.

[0017] Furthermore, the composition of the corrosion-resistant multi-principal alloy coating prepared in step S3 is Ti 34 Zr 15 Nb 24 Mo 4 V 23 、Ti 44 Zr 15 Nb 28 Mo 3 V 10 、Ti35 Zr 14 Nb 28 Mo 7 V 16 or Ti 22 Zr 11 Nb 23 Mo 16 V 28 .

[0018] Furthermore, the electrochemical corrosion performance indicators include corrosion potential and pitting potential.

[0019] Furthermore, the calculation method of the Pearson correlation coefficient is:

[0020]

[0021] In the formula, x i for and i are all variables of input features, and are the sample means of the corresponding variables.

[0022] Furthermore, these target prediction models are evaluated using the test set, and the model with the best performance is selected as the final target prediction model, including:

[0023] The test set is divided into different proportions, and the determination coefficient and root mean square error of the target prediction model are evaluated using the test set in each proportion, and the target prediction model with the best performance is selected.

[0024] Furthermore, the multi-principal component alloy coating is a high entropy alloy coating.

[0025] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: The design method of the corrosion-resistant multi-principal component alloy coating based on the machine learning model provided by the present invention combines the machine learning algorithm with the magnetron sputtering technology to develop a new multi-principal component alloy coating with high corrosion resistance, and can also effectively guide the alloy design and preparation process, thereby shortening the development cycle of new materials. The use of genetic algorithms for multi-objective optimization further guides the composition design and preparation parameter optimization, reducing the cost of new material development. The combination of machine learning and magnetron sputtering methods effectively accelerates the design and application of high-performance multi-principal component alloy coatings. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0027] Figure 1 A machine learning design flow chart of a corrosion-resistant multi-principal alloy coating according to an embodiment of the present invention;

[0028] Figure 2 It is a Pearson correlation heat map distribution diagram and feature importance ranking diagram between the components and process parameter characteristics of the embodiments of the present invention;

[0029] Figure 3 To rank the importance of process parameter characteristics targeting pitting potential;

[0030] Figure 4 Ranking of importance of process parameter characteristics targeting corrosion potential;

[0031] Figure 5 is the prediction performance of the random forest machine learning model of the embodiment of the present invention, wherein (a) and (b) are the E obtained by the random forest machine learning model respectively. pit and E corr Result;

[0032] Figure 6 1 and 1 , which are electrochemical test results of an embodiment of the present invention, wherein (a) is a potentiodynamic polarization curve in a 3.5 wt. % NaCl solution, and (b) is an impedance diagram in a 3.5 wt. % NaCl solution. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] The purpose of the present invention is to provide a design method for corrosion-resistant multi-principal alloy coatings based on a machine learning model, combine the machine learning algorithm with magnetron sputtering technology, guide the design and preparation process of corrosion-resistant multi-principal alloy coatings, and shorten the development cycle of new materials.

[0035] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0036] like Figure 1 As shown, the design method of the corrosion-resistant multi-principal alloy coating based on the machine learning model provided by the embodiment of the present invention includes the following steps:

[0037] S1, constructing a corrosion data set of a multi-principal alloy coating in 3.5wt.% NaCl and preprocessing it, wherein the corrosion data set includes alloy composition, magnetron sputtering process parameters and electrochemical corrosion performance indicators of the coating, and the preprocessed corrosion data set is divided into a training set and a test set;

[0038] S2, in Python's scikit-learn library, uses a variety of classic machine learning models to fit and model the training set, and obtains multiple target prediction models after training; then uses the test set to evaluate these target prediction models, and selects the model with the best performance as the final target prediction model;

[0039] S3, based on the final target prediction model, use a genetic algorithm to search for the optimal solution in the multi-target space to determine the optimal element content and magnetron sputtering process parameters, and prepare a multi-principal alloy coating according to the optimal element content and process parameters.

[0040] In this embodiment, the constructed data set is a corrosion data set of a multi-principal alloy coating prepared by magnetron sputtering in 3.5wt.% NaCl, and a total of 52 data sets are collected, each of which includes alloy composition, magnetron sputtering process parameters and coating corrosion resistance information. The process parameters in the data set include sputtering power (P), bias voltage (Bia), substrate temperature (T) and sputtering time (t), ranging from 100W to 2000W, -400V to 0V, 25℃ to 500℃ and 10min to 480min.

[0041] In this embodiment, the preprocessing includes: performing Pearson correlation coefficient analysis on the corrosion data set and drawing a heat map to evaluate the correlation between component characteristics and process parameter characteristics, and combining feature importance ranking to eliminate features with correlation coefficients greater than 0.9.

[0042] In this embodiment, in S2, multiple target prediction models are obtained through training, specifically including: based on the training set, taking the composition ratio of alloy elements and process parameters as input variables, and taking the electrochemical corrosion performance index as output variable, the selected target prediction model is trained to obtain a trained target prediction model.

[0043] In this embodiment, the magnetron sputtering process parameters include sputtering power (P), bias voltage (Bia), substrate temperature (T) and sputtering time (t).

[0044] In this embodiment, the electrochemical corrosion performance indicators include corrosion potential and pitting potential (E corr and E pit ).

[0045] In this embodiment, the target prediction models are evaluated using a test set, and the model with the best performance is selected as the final target prediction model, which specifically includes:

[0046] The test set is divided into different proportions, and the determination coefficient and root mean square error of the target prediction model are evaluated using the test set in each proportion, and the target prediction model with the best performance is selected.

[0047] like Figure 2-4 As shown in Figure 2, using Pearson correlation coefficient analysis can eliminate features with strong linear correlation (i.e., Pearson correlation coefficient ≥ 0.9), thereby simplifying the model, reducing computational costs, and improving the generalization ability of the model. Figure 2 In the matrix, each sector and corresponding square represents the correlation coefficient between two different variables. The Pearson correlation coefficient between Fe and Co is as high as 0.95. Given that the sample data for Fe is more evenly distributed, the Co feature was removed. For process parameters, the correlation coefficient between t and P reached 0.9. In order to reasonably exclude the process parameters with lower rankings in strong linear correlations, the magnetron sputtering process parameters were ranked according to their feature importance ( Figure 3 and Figure 4 ). The results show that for E pit and E corr , P is more important than t, so t is no longer used as an input feature. The calculation method of the Pearson correlation coefficient (p) is as follows:

[0048]

[0049] In the formula, x i for and i is the variable of input feature, and is the sample mean of the corresponding variable.

[0050] In this embodiment, in order to evaluate the performance of the machine learning algorithm, the original data set is randomly divided into a training set and a test set, and the proportion of the test set is gradually increased from 10% to 50%.

[0051] In this step, the coefficient of determination (R 2 ) and root mean square error (RMSE) to evaluate the model accuracy, and the calculation formula is:

[0052]

[0053] Among them, yi is the sample experimental data, Represents the model prediction result, is the average value of the sample experimental data; and finally 20% of the test set is determined to be the best ratio.

[0054] In this embodiment, there are five kinds of classic machine learning models, including random forest (RF), gradient boosting regression (GBR), artificial neural network (ANN), support vector regression (SVR) and extreme gradient boosting (XGBoost) models. In step S2, the target prediction model with the highest accuracy is saved. When the test set ratio is 20%, the RF model predicts E pit and E corr The best combination of performance.

[0055] Calculate RMSE and R of 5 machine learning models using the multi-holdout method 2 , to evaluate the accuracy of the model. RMSE and R were obtained by repeating the calculation 500 times. 2 The average value of , finally choose RF machine learning algorithm. Prediction E pit The R2 / RMSE values ​​in the training set are 0.945±0.01 / 0.123±0.01 (V SCE ), and 0.799±0.07 / 0.236±0.06 (V SCE ). Prediction E corr R 2 / RMSE values ​​in the training set and test set were 0.944±0.02 / 0.059±0.01 (V SCE ) and 0.760±0.08 / 0.148±0.05(V SCE ).

[0056] E obtained by RF model pit and E corr Prediction Figure 5 The experimental values ​​and predicted values ​​are distributed along the diagonal, indicating that the machine learning model has good generalization ability. pit and E corr The corresponding R 2 The values ​​are 0.87 and 0.86, respectively, which can be regarded as indicators of robust regression results. Therefore, the prediction results of the RF model can be used as a prediction model for subsequent optimization of target performance.

[0057] In this embodiment, in step S4, a composition of Ti is obtained through four iterations. 34 Zr 15 Nb 24 Mo 4 V 23 、Ti44 Zr 15 Nb 28 Mo 3 V 10 、Ti 35 Zr 14 Nb 28 Mo 7 V 16 and Ti 22 Zr 11 Nb 23 Mo 16 V 28 Multi-principal alloy coating.

[0058] In this embodiment, the multi-principal alloy coating is a high entropy alloy coating.

[0059] Based on the optimized composition and process parameters, a multi-principal alloy coating was prepared, and the corrosion resistance of the coating was tested and verified. The mechanism of the coating's excellent corrosion resistance was analyzed by combining electrochemical and microstructural experiments. Figure 6 As shown in the electrochemical corrosion test results, Ti 35 Zr 14 Nb 28 Mo 7 V 16 The multi-element alloy coating exhibits the best corrosion resistance. pit 1931.1mV SCE , the passivation region is 1917.3mV SCE Compared with the existing multi-principal alloy coatings in the dataset, Ti 35 Zr 14 Nb 28 Mo 7 V 16 The passivation performance of the multi-element alloy coating was improved by 15%, and the pitting corrosion resistance was improved by 23.6%. 35 Zr 14 Nb 28 Mo 7 V 16 The polarization resistance value of the multi-principal alloy coating is 2.3 times that of 316L stainless steel, which indicates that the coating prepared by the method of the present invention greatly enhances the corrosion resistance of the 316L stainless steel substrate.

[0060] In summary, the present invention provides a design method for corrosion-resistant multi-principal alloy coatings based on a machine learning model. The present invention designs new materials through a machine learning method, and achieves the purpose of rapidly designing high-performance multi-principal alloy coatings through random forest modeling and genetic algorithm optimization, while guiding the optimization of alloy composition and process parameters. Moreover, the machine learning-accelerated design method for high-corrosion-resistant multi-principal alloy coatings proposed by the present invention replaces the design strategy of traditional alloys, avoids a large number of repeated experiments, and reduces the research and development cycle and cost of new materials. In addition, the Ti designed by the present invention 35 Zr 14 Nb 28 Mo 7 V 16 Compared with the existing multi-principal alloy coatings in the data set, the multi-principal alloy coating has a 15% higher passivation performance and a 23.6% higher pitting corrosion resistance, achieving an integrated design of composition, process and performance.

[0061] The remaining technical features in this embodiment can be flexibly selected by those skilled in the art according to actual conditions to meet different specific practical needs. However, it is obvious to those skilled in the art that it is not necessary to adopt these specific details to implement the present invention. In other examples, in order to avoid confusing the present invention, the well-known components, structures or parts are not specifically described, which are all within the technical protection scope defined by the technical solution claimed for protection in the claims of the present invention.

[0062] Modifications and changes made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the scope of protection of the claims attached to the present invention. In the above description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it is obvious to those of ordinary skill in the art that these specific details are not necessary to practice the present invention. In other examples, in order to avoid confusing the present invention, well-known technologies, such as specific construction details, operating conditions and other technical conditions, are not specifically described.

[0063] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A design method for corrosion-resistant multi-principal alloy coating based on a machine learning model, characterized in that: The following steps are involved: S1, constructing a corrosion data set of a multi-principal alloy coating in 3.5wt.% NaCl and preprocessing it, wherein the corrosion data set includes alloy composition, magnetron sputtering process parameters and electrochemical corrosion performance indicators of the coating, and the preprocessed corrosion data set is divided into a training set and a test set; S2, in Python's scikit-learn library, uses a variety of classic machine learning models to fit and model the training set, and obtains multiple target prediction models after training; then uses the test set to evaluate these target prediction models, and selects the model with the best performance as the final target prediction model; S3, based on the final target prediction model, use a genetic algorithm to search for the optimal solution in the multi-target space to determine the optimal element content and magnetron sputtering process parameters, and prepare a multi-principal alloy coating according to the optimal element content and process parameters.

2. The design method of corrosion-resistant multi-principal alloy coating based on machine learning model according to claim 1 is characterized in that: The preprocessing includes: performing Pearson correlation coefficient analysis on the corrosion data set and drawing a heat map to evaluate the correlation between component characteristics and process parameter characteristics, and combining feature importance sorting to eliminate features with a correlation coefficient greater than 0.

9.

3. The design method of corrosion-resistant multi-principal alloy coating based on machine learning model according to claim 1 is characterized in that: In S2, multiple target prediction models are obtained through training, specifically including: based on the training set, taking the composition ratio of alloy elements and process parameters as input variables, and taking the electrochemical corrosion performance index as output variable, the selected target prediction model is trained to obtain a trained target prediction model.

4. The design method of corrosion-resistant multi-principal alloy coating based on machine learning model according to claim 1 is characterized in that: There are five kinds of classic machine learning models, including random forest, gradient boosting regression, artificial neural network, support vector regression and extreme gradient boosting model.

5. The design method of corrosion-resistant multi-principal alloy coating based on machine learning model according to claim 1, characterized in that: The magnetron sputtering process parameters include sputtering power, bias voltage, substrate temperature and sputtering time.

6. The design method of corrosion-resistant multi-principal alloy coating based on machine learning model according to claim 1, characterized in that: The composition of the corrosion-resistant multi-principal alloy coating prepared in step S3 is Ti 34 Zr 15 Nb 24 VxD 23 、Ti 44 Zr 15 Nb 28 V3V 10 、Ti 35 Zr 14 Nb 28 VxF 16 or Ti 22 Zr 11 Nb 23 Mo 16 V 28 .

7. The design method of corrosion-resistant multi-principal alloy coating based on machine learning model according to claim 1, characterized in that: The electrochemical corrosion performance indicators include corrosion potential and pitting potential.

8. The design method of corrosion-resistant multi-principal alloy coating based on machine learning model according to claim 2, characterized in that: The calculation method of the Pearson correlation coefficient is: In the formula, x i and i are all variables of input features, and are the sample means of the corresponding variables.

9. The design method of corrosion-resistant multi-principal alloy coating based on machine learning model according to claim 1, characterized in that: In S2, the target prediction models are evaluated using the test set, and the model with the best performance is selected as the final target prediction model, which specifically includes: The test set is divided into different proportions, and the determination coefficient and root mean square error of the target prediction model are evaluated using the test set in each proportion, and the target prediction model with the best performance is selected.

10. The design method of corrosion-resistant multi-principal alloy coating based on machine learning model according to claim 1, characterized in that: The multi-principal-component alloy coating is a high-entropy alloy coating.

Images